Respiratory Effort Belts in Postoperative Respiratory Monitoring:

Pilot Study with Different Patients

Tiina M Seppänen

1, 2

, Olli-Pekka Alho

2, 3, 4

, Merja Vakkala

2, 5

, Seppo Alahuhta

2, 5

and Tapio Seppänen

1, 2

Center for Machine Vision and Signal Analysis, University of Oulu, P.O Box 4500, 90014 Oulu, Finland

Medical Research Center Oulu, Oulu University Hospital and University of Oulu, Oulu, Finland

Department of Otorhinolaryngology, Oulu University Hospital, Oulu, Finland

Research Unit of Otorhinolaryngology and Ophthalmology, University of Oulu, Oulu, Finland

Department of Anesthesiology, Oulu University Hospital, Oulu, Finland

Keywords: Airflow Waveform, Respiratory Airflow, Respiratory Rate, Respiratory Volume.

Abstract: Respiratory complications are common in patients after the general anaesthesia. Respiratory depression often

occurs in association with postoperative opioid analgesia. Currently, there is a need for a continuous non-

invasive respiratory monitoring of spontaneously breathing postoperative patients. We used calibrated

respiratory effort belts for the respiratory monitoring pre- and postoperatively. Used calibration method

enables accurate estimates of the respiratory airflow waveforms. Five different patients were measured with

the spirometer and respiratory effort belts at the same time. Preoperative measurements were done in the

operating room just before the operation, whereas postoperative measurements were done in the recovery

room after the operation. We compared three calibration models pre- and postoperatively. Postoperative

calibration produced more accurate respiratory airflows. Results show that not only the tidal volume, minute

volume and respiratory rate can be computed precisely from the estimated respiratory airflow, but also the

respiratory airflow waveforms are very accurate. The method produced accurate estimates even from the

following challenging respiratory signals: low airflows, COPD, hypopneic events and thoracoabdominal

asynchrony. The presented method is able to produce estimates of postoperative respiratory airflow

waveforms to enable accurate, continuous, non-invasive respiratory monitoring postoperatively.

1 INTRODUCTION

Respiratory complications are common in patients

after the general anaesthesia. Respiratory depression

often occurs in association with postoperative opioid

analgesia (Etches, 1994; Gamil and Fanning, 1991;

Taylor et al., 2005). Adequate respiration monitoring

postoperatively is important, so that respiratory

depression can be identified as early as possible

(George et al., 2010; Lynn and Curry, 2011).

During general anaesthesia, mechanical

ventilation is used and, consequently, monitoring of

gas exchange and respiration can be done accurately.

During postoperative care, respiratory status can be

assessed with oxygen saturation (SpO

)

measurements, blood gas measurements, subjective

clinical assessment and intermittent, manual

measurements of respiratory rate (Lynn and Curry,

2011; Ramsay et al., 2013). The problems with these

methods are that they are slow and, in addition,

subjective methods are also unreliable and give

inconsistent results (Lovett, 2005). There is a need for

a continuous non-invasive respiratory monitoring of

spontaneously breathing postoperative patients.

Respiratory depression and subsequent adverse

outcomes can arise from pain, residual operating

room anaesthetics and administration of opioids for

pain management (Cepeda et al., 2003). Inadequate

respiration can result in respiratory complications,

morbidity, mortality and excessive costs.

A few studies have been recently published on

monitoring postoperative respiration continuously

and non-invasively. Drummond et al., (2013) have

studied respiratory rate and breathing patterns of

postoperative subjects using encapsulated tri-axial

accelometer taped to a subject’s body. They found

that abnormal breathing patterns are extremely

common. Voscopoulos et al., (2015; 2014a; 2014b)

Seppänen, T., Alho, O-P., Vakkala, M., Alahuhta, S. and Seppänen, T.

Respiratory Effort Belts in Postoperative Respiratory Monitoring: Pilot Study with Different Patients.

DOI: 10.5220/0005695000760082

In Proceedings of the 9th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2016) - Volume 5: HEALTHINF, pages 76-82

ISBN: 978-989-758-170-0

have studied minute ventilation, tidal volume and

respiratory rate of postoperative subjects using

impedance-based electrodes placed to a subject’s

body.

Recently, we published a novel calibration

method to produce accurate estimates of respiratory

airflow signals from respiratory effort belt signals

(Seppänen, 2013). Here, the method is used in order

to produce estimates of postoperative respiratory

airflow waveforms to enable accurate, continuous,

non-invasive respiratory monitoring postoperatively.

Pre- and postoperative measurement data of different

patients are used to demonstrate the performance of

the method.

2 MATERIALS AND METHODS

2.1 Materials

The study protocol was approved by the Regional

Ethics Committee of the Northern Ostrobothnia

Hospital District. Five patients who had lumbar back

surgery and were expected to need opioid analgesia

postoperatively were recruited to the study. Exclusion

criteria were BMI (Body Mass Index) over 40 and the

planned surgical wound being in the area where

respiratory effort belts were placed. The

characteristics of the volunteers are given in Table 1.

Table 1: Characteristics of volunteers.

Patient Gender

Age

[years]

BMI

[kg/m

]

Disease

1 M 68 21.8 None

2 M 41 30.3 None

3 F 77 22.4 None

4 M 64 28.1 COPD

5 M 67 27.4 Sleep apnea

Respiratory effort belt signals were recorded with

the polygraphic recorder (Embletta Gold, Denver,

Colorado, USA). The recorder had inductive

respiratory effort belts for rib cage and abdomen with

the sampling rate of 50 Hz. For calibrating the

respiratory effort belt signals, simultaneous

respiratory airflow signal was recorded with a

spirometer (Medikro Pro M915, Medikro Oy,

Kuopio, Finland), which had a sampling rate of 100

Hz. Mask (Cortex Personal-Use-Mask, Leipzig,

Germany) was attached to the mouthpiece of the

spirometer. The spirometer could record at most 1

min long signals.

2.2 Measurement Protocol

The measurements for each patient were done in two

parts: 1) short measurement session (5 min)

preoperatively; and 2) longer measurement session (3

h) postoperatively.

The first measurements were done in the

operating room just before the operation without any

sedative premedication. The rib cage respiratory

effort belt was placed on the xyphoid process and the

abdominal belt was placed above the umbilicus. The

mask of the spirometer was put to the volunteer’s face

and its airtightness was secured. The signals were

recorded until two successful recordings of the 1 min

were obtained. The places of respiratory effort belts

were marked with drawing ink on the skin, so that it

was possible to place the belts to the same places

postoperatively. The mask and respiratory effort belts

were removed.

As soon as it was possible, measurements were

continued postoperatively in the recovery room. The

rib cage respiratory effort belt and the abdominal

respiratory effort belt were placed to the previously

marked places. They recorded the signals during the

whole 3 hour measurement period. Every 10 min, the

mask of the spirometer was put to the volunteer’s

face, its airtightness was secured and 1 min

measurement with the spirometer was recorded.

Participation to the study did not affect the routine

management of the patients.

2.3 Calculation of Respiratory Airflow

Estimates

In this study, we applied our recently published

respiratory effort belt calibration method (Seppänen,

2013). The method was therein tested against various

breathing style changes and body position changes,

and compared with the state-of-the-art methods. Our

method outperformed the other methods showing

high robustness to the breathing style changes and

body position changes.

Our method is an extension to conventional

multiple linear regression method so that 1) it uses

number N of consecutive input signal samples and

linear filtering for estimation of each output signal

sample and 2) it is based on polynomial regression to

model different transfer functions between the input

and output. The method is based on optimally trained

FIR (Finite Impulse Response) filter bank constructed

as a MISO (Multiple-Input Single-Output) system

between the respiratory effort belt signals and the

spirometer signal. Three polynomial transfer

functions were tested: linear terms only (M1), linear

Respiratory Effort Belts in Postoperative Respiratory Monitoring: Pilot Study with Different Patients

terms and cross-product term (M2), and linear terms

with second order terms (M3).

Figure 1 and equation 1 show the realization of the

filter bank for model M2. Similar realizations can be

derived also for models M1 and M3.















































































(1)

where 





, 





and 





denote the N tap coefficients of

filters FIR

, FIR

and FIR

, respectively: 















,







,…,









, i = 1,2,3. Superscript T

denotes matrix transpose. Parameter  denotes

respiratory airflow from spirometer and  is zero-

mean Gaussian error. Vectors 



and 



include N

consecutive signal samples from the rib cage

respiratory effort belt signal and abdominal

respiratory effort belt signal, respectively: 





















,





1



,…,





1







, j = 1,2 and k

= N,…,n. Variable n is the number of observations

used in the calibration.

Figure 1: Extended calibration method of respiratory effort

belt signals.

 is an (n-N+1) × (3 × N) matrix formed from the

vectors 



and 



















…



















…





























…



















.

(2)

During the calibration, tap coefficients





, 





and 





are estimated with the method of least-squares. The

least-squares estimator of the parameter vector 











,





,









is given by























 .

(3)

The length of the vector 



is 3 × N. Finally, the

respiratory airflow signal estimated from the rib cage

and abdominal respiratory effort belt signals through

the FIR filter bank is









(4)

In Figure 1, there is the delay element z

-D

included at

the output. There is always a small delay between the

spirometer signal and respiratory effort belt signals

due to the physiological reasons and internal delays

of measuring devices. Thus, the signals have to be

time-synchronized by searching for a proper value for

D (Seppänen, 2013).

We made two different test setups: 1) the data of

the second preoperative measurement minute were

used to train the estimation model and the data of all

the postoperative measurement minutes were used to

test the estimation model (PRE setup); and 2) the data

of the first postoperative measurement minute were

used to train the estimation model and data of the rest

postoperative measurement minutes were used to test

the estimation model (POST setup).

The similarity of spirometer signals and estimated

respiratory airflow signals were assessed by

computing R

(coefficient of determination) values.

Tidal volumes, minute volumes and BPM (Breaths

per Minute, respiratory rate) were calculated from the

spirometer signals and estimated respiratory airflow

signals. Relative errors were calculated.

3 RESULTS AND DISCUSSION

Signals were recorded according to the protocol

described in Section 2.2. There were altogether 93

simultaneous measurement minutes with spirometer

and respiratory effort belts. Five of these had to be

discarded due to a malfunction of the spirometer. One

of them had to be discarded due to a malfunction of

the polygraphic recorder. In addition to that, five

postoperative measurement minutes of patient 5 had

to be discarded, because he had serious difficulties to

wake up and stay awake in the recovery room. During

the measurements patients received opioid analgesia

as many times as they needed: 3, 2, 1, 7 and 4 times

for patients 1-5, respectively.

During the measurements a number of problems

related to PRE setup were observed. Firstly, places of

respiratory effort belts can interchange before POST

setup by mistake. Secondly, there may be a need to

tighten or loosen the respiratory effort belts after the

operation, because fluids can accumulate in the body

or can leave from the body during the operation. This

leads to a situation where the estimation model

trained with a preoperative data is not valid anymore.

Thirdly, if there are complications during the

operation or if the operation is prolonged otherwise,

the estimation model trained with a preoperative data

can be erroneous for the postoperative data. With the

POST setup, no problems were observed.

HEALTHINF 2016 - 9th International Conference on Health Informatics

3.1 Accuracy of Airflow Estimates

The selection of the best model and FIR filter length

(N value) depended on whether waveform accuracy

), tidal volume (V

) error, minute volume (V

minute

)

error or BPM error values were studied. Table 2

summarizes the results when PRE setup was used for

training and testing all estimation models. It is seen in

Table 2 that model M1 produced the best results.

Table 2: Results (average value ± SD) with the best models

when PRE setup was used for training and testing all

estimation models.

Best model M1 M1 M1 M1

FIR size N = 8 N = 8 N = 8 N = 16

Patient R

Abs (V

error) [%]

Abs

minute

error) [%]

BPM error

1 0.88 ± 0.05 12.8 ± 10.7 12.5 ± 9.5 0.03 ± 0.07

2 0.87 ± 0.06 26.7 ± 7.7 25.1 ± 8.2 0.01 ± 0.01

3 0.94 ± 0.02 10.9 ± 2.9 14.2 ± 5.0 0.09 ± 0.30

4 0.87 ± 0.06 15.9 ± 6.0 21.7 ± 8.6 0.09 ± 1.14

5 0.43 ± 0.17 70.2 ± 14.1 72.4 ± 12.3 0.48 ± 0.79

Average 0.82 ± 0.18 24.1 ± 21.4 26.0 ± 21.3 0.29 ± 0.70

Table 3 summarizes the results when POST setup

was used for training and testing all estimation

models. In this case, it is seen that model M2

produced the best results with this data. As an

important difference to the preceding results in Table

2, more accurate waveforms were received since R

values were higher and BPM error values lower.

Also, the volume error values decreased to the

fractions.

Table 3: Results (average value ± SD) with the best model

when POST setup was used for training and testing all

estimation models.

Best model M2 M2 M2 M2

FIR size N = 8 N = 16 N = 16 N = 8

Patient R

Abs (V

error) [%]

Abs

minute

error) [%]

BPM error

1 0.90 ± 0.04 11.4 ± 7.9 9.9 ± 6.8 0.01 ± 0.01

2 0.95 ± 0.01 5.9 ± 4.5 5.8 ± 5.0 0.01 ± 0.01

3 0.94 ± 0.04 5.4 ± 4.7 6.3 ± 4.2 0.10 ± 0.30

4 0.88 ± 0.09 8.7 ± 6.0 11.9 ± 10.2 0.83 ± 1.13

5 0.91 ± 0.02 10.5 ± 9.3 8.2 ± 6.3 0.01 ± 0.00

Average 0.91 ± 0.06 5.8 ± 6.3 8.5 ± 7.1 0.21 ± 0.63

It can be seen from Table 2 and Table 3, that if the

smallest error results are sought, then there is a need

to use several models. However, if there is, for

example, a need to get accurate respiratory rate only,

then one should choose POST setup and model M2

with N value 8. On the other hand, one may wish to

use only one model with good overall performance.

The results for that are presented next. Model M3

(including linear and 2

order terms) produced

clearly worse results than the other models, thus only

the results of using models M1 and M2 are presented

here. Table 4 and Table 5 present the results when the

estimation model M1 with N value 8 was used with

PRE and POST setups, respectively.

Table 4: Results (average value ± SD) of the calibration

when estimation model M1 (N = 8) was used with PRE

setup.

Patient R

Abs (V

error) [%]

Abs (V

minute

error) [%]

BPM error

1 0.88 ± 0.05 12.8 ± 10.7 12.5 ± 9.5 0.09 ± 0.27

2 0.87 ± 0.06 26.7 ± 7.7 25.1 ± 8.2 0.01 ± 0.01

3 0.94 ± 0.02 10.9 ± 2.9 14.2 ± 5.0 0.09 ± 0.29

4 0.87 ± 0.06 15.9 ± 6.0 21.7 ± 8.6 1.00 ± 1.06

5 0.43 ± 0.17 70.2 ± 14.1 72.4 ± 12.3 0.37 ± 0.55

Average 0.78 ± 0.20 24.1 ± 21.4 26.0 ± 21.3 0.32 ± 0.68

Table 6 and Table 7 present the results when the

estimation model M2 with N value 8 was used with

PRE and POST setups, respectively.

It is clearly seen from Tables 4-7 that POST setup

produced superior results. Respiratory airflow

waveforms are much more accurate, average R

increased from 0.78 to 0.91 with both models M1 and

M2. In addition to that, tidal volume errors, minute

volume errors and BPM errors are smaller. However,

when the average results of Table 5 and Table 7 are

compared, it can be seen that models M1 and M2 with

N value 8 produced both very good results and that

there are little differences between the results.

Table 5: Results (average value ± SD) of the calibration

when estimation model M1 (N = 8) was used with POST

setup.

Patient R

Abs (V

error) [%]

Abs (V

minute

error) [%]

BPM error

1 0.91 ± 0.04 11.6 ± 7.8 8.9 ± 7.1 0.01 ± 0.01

2 0.94 ± 0.02 6.9 ± 7.0 5.7 ± 5.6 0.01 ± 0.01

3 0.94 ± 0.04 5.6 ± 4.5 6.0 ± 3.8 0.10 ± 0.30

4 0.87 ± 0.09 8.4 ± 4.9 11.9 ± 8.9 0.82 ± 1.13

5 0.90 ± 0.02 9.6 ± 5.3 10.9 ± 4.9 0.12 ± 0.35

Average 0.91 ± 0.06 8.4 ± 6.1 8.6 ± 6.7 0.23 ± 0.64

Table 6: Results (average value ± SD) of the calibration

when estimation model M2 (N = 8) was used with PRE

setup.

Patient R

Abs (V

error) [%]

Abs

minute

error) [%]

BPM error

1 0.82 ± 0.08 17.8 ± 9.8 21.3 ± 8.3 0.10 ± 0.27

2 0.76 ± 0.09 42.7 ± 8.8 51.7 ± 9.6 0.04 ± 0.12

3 0.93 ± 0.03 13.3 ± 5.1 15.3 ± 6.4 0.09 ± 0.30

4 0.87 ± 0.06 15.4 ± 6.7 19.6 ± 8.6 0.91 ± 1.11

5 0.39 ± 0.21 55.1 ± 9.5 65.5 ± 15.1 1.67 ± 1.47

Average 0.78 ± 0.20 26.9 ± 17.8 32.4 ± 21.1 0.49 ± 0.96

Respiratory Effort Belts in Postoperative Respiratory Monitoring: Pilot Study with Different Patients

3.2 Example Cases

Figure 2 depicts short segment of example signals

with low airflow. Estimation model M2 (N = 16) was

used with the POST setup for the measurement

signals of patient 3. In this case, R

was 0.94, tidal

volume error 1.5 %, minute volume error 10.9 % and

BPM error 0.01. The spirometer signal is estimated

with excellent accuracy.

Figure 3 depicts an example of apneic event. In

this case, estimation model M1 (N = 8) was used with

the POST setup for the measurement signals of

patient 5. During obstruction, the rib cage ceases to

move, but abdomen is moving. There is no air

exchange, so there is no airflow signal either. It can

be seen from the figure, that during the obstruction

airflow was zero, but because there was movement in

respiratory effort belts, the estimated respiratory

airflow signal also showed activity. The same

phenomenon was also encountered by Drummond et

al., (2013) and Voscopoulos et al., (2013). However,

apneic events can be detected from the changed

pattern of movements from respiratory effort belt

signals. Both of them have lower amplitudes than

before or after the apneic event. Especially the rib

cage respiratory effort belt is almost motionless

during the apneic event. Detected parts of the

respiratory airflow estimate could then be replaced

with zero airflow.

Table 7: Results (average value ± SD) of the calibration

when estimation model M2 (N = 8 was used with POST

setup.

Patient R

Abs (V

error) [%]

Abs (V

minute

error) [%]

BPM error

1 0.90 ± 0.04 11.4 ± 7.6 10.0 ± 7.0 0.01 ± 0.01

2 0.95 ± 0.01 6.1 ± 4.8 6.2 ± 5.3 0.01 ± 0.01

3 0.94 ± 0.04 5.3 ± 4.5 6.2 ± 4.0 0.10 ± 0.30

4 0.88 ± 0.09 8.8 ± 3.9 11.4 ± 10.1 0.83 ± 1.13

5 0.91 ± 0.02 13.1 ± 8.9 9.6 ± 6.4 0.01 ± 0.00

Average 0.91 ± 0.06 8.8 ± 6.6 8.7 ± 7.1 0.21 ± 0.63

Figure 2: Short segment of example signals with low

airflow: spirometer signal (thin line) and the estimated

respiratory airflow signal (bold line).

Figure 4 depicts hypopneic event of patient 4 with

COPD. Here, estimation model M2 (N = 8) was used

with the POST setup. In this case, R

was 0.81, tidal

volume error was 0.8 %, minute volume error was

3.24 % and BMP error was 1.08. It can be seen, that

the method was able to estimate respiratory airflow

very well even in this kind of complicated situations.

Figure 3: Short segment of example signals during apneic

event. Upper subfigure: spirometer signal (solid line) and

the estimated respiratory airflow signal (dotted line). Lower

subfigure: rib cage respiratory effort belt signal (solid line)

and abdominal respiratory effort belt signal (dotted line).

Figure 4: Example signals of COPD patient. Upper

subfigure: spirometer signal (solid line) and the estimated

respiratory airflow signal (dotted line). Lower subfigure: rib

cage respiratory effort belt signal (solid line) and abdominal

respiratory effort belt signal (dotted line).

Measurement data of patient 4 included

thoracoabdominal asynchrony more or less during the

whole measurement session. Figure 5 depicts one

example of this. Estimation model M2 (N = 8) was

used with the POST setup and the results were the

following: R

was 0.94, tidal volume error was 4.7 %,

minute volume error was 2.6 % and BPM error was

0.02. These results are consistent with our earlier

findings indicating that our method produces very

good results with thoracoabdominal asynchrony

signals too (Seppänen, 2013).

HEALTHINF 2016 - 9th International Conference on Health Informatics

Figure 6 demonstrates the performance difference

between PRE and POST setups. In this case,

estimation model M1 (N = 16) was used firstly with

the PRE setup and secondly with the POST setup. The

measurement data was from patient 1. It can be seen

from Figure 6 that there were clear differences with

the estimated respiratory airflows (PRE setup with

dotted line and POST setup with bold line). Following

numerical results demonstrate these differences

further. The results for the PRE setup were: R

was

0.82, tidal volume error was 34.8 %, minute volume

error was 35.8 % and BPM error was 0.00.

Equivalently results for the POST setup were: R

was

0.93, tidal volume error was 14.6 %, minute volume

error was 13.9 % and BPM error was 0.00. Although,

the PRE setup was otherwise remarkably worse than

POST setup in this case, BPM was estimated very

accurately.

Figure 5: Example signals of thoracoabdominal

asynchrony. Upper subfigure: spirometer signal (solid line)

and the estimated respiratory airflow signal (dotted line).

Lower subfigure: rib cage respiratory effort belt signal

(solid line) and abdominal respiratory effort belt signal

(dotted line).

Figure 6: Short segment of example signals depicting the

difference of PRE and POST setups: spirometer signal (thin

line), the first estimated respiratory airflow signal (PRE

setup, dotted line) and the second estimated respiratory

airflow (POST setup, bold line).

3.3 Limitations of Study

The study included a number of limitations. Firstly,

the study included only five patients. The study

should be repeated with a larger data set in order to

draw more general conclusions. Secondly, respiratory

effort belts cannot be used if the surgical wound is in

the area where the belts are placed. However, the

proposed method could be applied to the

measurement data acquired with other sensors

without this kind of restriction, such as acceleration

sensors. Thirdly, as was pointed out in Section 3.2

during the apneic event there is no respiratory airflow

but still estimated respiratory airflow shows

otherwise. This could be resolved by detecting the

changed pattern of movements from respiratory effort

belts and replacing these parts of the respiratory

airflow estimate with zero airflow. This remains

future work.

4 CONCLUSIONS

Here, a method was proposed to estimate accurate

continuous respiratory airflow postoperatively. The

data from respiratory effort belts were calibrated with

a spirometer using an extended multiple linear

regression method. The results showed that training

the estimation model with the postoperative data

produced much more accurate results than training

the estimation model with the preoperative data.

It was demonstrated with data from five different

patients in postoperative situation that estimated

respiratory airflow signals have very accurate

waveforms. In addition, tidal volume, minute volume

and respiratory rate can be calculated remarkably

accurately from these signals. The method produced

very good estimates even from challenging

respiration signals: low airflows, COPD, hypopneic

events and thoracoabdominal asynchrony.

In summary, the presented method is able to

produce estimates of postoperative respiratory

airflow waveforms to enable accurate, continuous,

non-invasive respiratory monitoring postoperatively.

ACKNOWLEDGEMENTS

Finnish Cultural Foundation, North Ostrobothnia

Regional Fund and International Doctoral

Programme in Biomedical Engineering and Medical

Physics (iBioMEP) are gratefully acknowledged for

financial support.

Respiratory Effort Belts in Postoperative Respiratory Monitoring: Pilot Study with Different Patients

REFERENCES

Cepeda M.S., Farrar J.T., Baumgarten M., Boston R., Carr

D.B., Strom B.L., 2003. Side effects of opioids during

short-term administration: Effect of age, gender, and

race. Clinical Pharmacology & Therapeutics, 74(2):

102-112.

Drummond G.B., Bates A., Mann J., Arvind D.K., 2013.

Characterization of breathing patterns during patient-

controlled opioid analgesia. British Journal of

Anaesthesia, 111(6): 971-978.

Etches R., 1994. Respiratory depression associated with

patient-controlled analgesia: a review of eight cases.

Canadian Journal of Anaesthesia, 41(2): 125-132.

Gamil M., Fanning A., 1991. The first 24 hours after

surgery. A study of complications after 2153

consecutive operations. Anaesthesia, 46: 712-715.

George J.A., Lin E.E., Hanna M.N., Murphy J.D., Kumar

K., Ko P.S., Wu C.L., 2010. The effect of intravenous

opioid patient-controlled analgesia with and without

background infusion on respiratory depression: a meta-

analysis. Journal of Opioid Management, 6(1): 47-54.

Lovett P.B., Buchwald J.M., Sturmann K., Bijur P., 2005.

The vexatious vital: neither clinical measurements by

nurses nor an electronic monitor provides accurate

measurements of respiratory rate in triage. Annals of

Emergency Medicine, 45(1): 68-76.

Lynn L.A., Curry J.P., 2011. Patterns of unexpected in-

hospital deaths: a root cause analysis. Patient Safety in

Surgery, 5:3.

Ramsay M.A.E., Usman M., Lagow E., Mendoza M.,

Untalan E., De Vol E., 2013. The accuracy, precision

and reliability of measuring ventilatory rate and

detecting ventilatory pause by rainbow acoustic

monitoring and capnometry. Anesthesia & Analgesia,

117(1): 69-75.

Seppänen, T.M., Alho, O.-P., Seppänen T., 2013. Reducing

the airflow waveform distortions from breathing style

and body position with improved calibration of

respiratory effort belts. Biomedical Engineering

Online, 12:97.

Taylor S., Kirton O.C., Staff I, Kozol R.A., 2005.

Postroperative day one: a high risk period for

respiratory events. The American Journal of Surgery,

190(5): 752-756.

Voscopoulos C.J., MacNabb C.M., Brayanov J., Qin L.,

Freeman J., Mullen G.J., Ladd D., George E., 2015. The

evaluation of a non-invasive respiratory volume

monitor in surgical patients undergoing elective surgery

with general anesthesia. Journal of Clinical Monitoring

and Computing, 29(2): 223-230.

Voscopoulos C., Ladd D., Campana L., George E., 2014.

Non-invasive respiratory volume monitoring to detect

apnea in post-operative patients: case series. Journal of

Clinical Medicine Research, 6(3): 209-214.

Voscopoulos C.J., MacNabb C.M., Freeman J., Galvagno

S.M., Ladd D., George E., 2014. Continuous

noninvasive respiratory volume monitoring for the

identification of patients at risk for opioid-induced

respiratory depression and obstructive breathing

patterns. Journal of Trauma and Acute Care Surgery,

77(3): S208-S215.

Voscopoulos C, Brayanov J., Ladd D., Lalli M., Panasyuk

A., Freeman J., 2013. Evaluation of a novel noninvasive

respiration monitor providing continuous measurement

of minute ventilation in ambulatory subjects in a variety

of clinical scenarios. Anesthesia & Analgesia, 117(1):

91-100.

HEALTHINF 2016 - 9th International Conference on Health Informatics